Electrophysiology
electrophysiology: using the electrical activity generated by the nervous system to study
neural computation, cognition, and behavior
→ cognition, emotions, etc. range from 1 to 100 Hz
How to measure neuroelectric signals
- intracellular recording: measuring from inside the neuron
- juxtacellular recording: measuring next to one specific cell in the membrane of the
brain
- extracellular recording: measuring spikes in the brain (one electrode)
- LFP/iEEG: measuring the sum of potentials in the brain (more electrodes)
- ECoG: measuring electric current on the dura (like EEG but the electrode is placed
under the skull (not in the brain))
- M/EEG: measurement outside of the head (non-invasive) → every magnetic field you
have has a perpendicular electric field
- MUA: measure collection of spikes (a little bit like LFP)
→ the further you get from the neuron, the weaker the potential that you measure with your
electrode
Advantages of electrophysiology
- cognition = fast → ephys captures this in the time-frame in which events occur (fMRI
cannot do this due to low temporal resolution) → direct measure of electrical brain
activity
- provides complex data → analyses of local regions and networks (fMRI is only one-
dimensional (up or down signal))
- link across scales/methods and species
→ species with different sizes of brains all have similar neural oscillations at almost
the same speeds (oscillations are constant across species)
Disadvantages of electrophysiology
- limited to large-scale potentials → field-cancellation might occur
- uncertainties in anatomical localization (low spatial resolution)
- data analysis is complicated, time-consuming, and annoying
- high temporal resolution → if you don’t exactly know what you are looking for you
might have a hard time finding it
How to analyze spiking data
problem: one electrode, many neurons
- use a low-pass spatial filter (LFP filter) to get local field potential
, - use spike filter → spikes contain information from many different neurons → use spike
sorting (assign each spike to a different neuron)
Traditional solution (to the aforementioned problem):
● intra/juxtacellular ephys → only measure one neuron (takes a long time to collect
data, and is ethically not doable)
Modern solution:
● spike-sorting (spike extraction → waveforms alignment → principal components
analysis → clustering (algorithm to separate spikes) → sorted waveforms)
- there are a lot of algorithms that solve this problem (hard to solve)
- spike-sorting leads to spikes going missing
, ● population → one neuron does not do exactly one thing, so we look at groups of
neurons that fire together
How to analyze LFP/MEEG data
- task-related vs. continuous
ERPs: event-related potentials → task-related analyses → take the average of all trials
- time-domain
- spectral domain
- time-frequency domain
Fourier transform: take a signal and take a sine wave as template and look at how similar
the signal is to the sine wave (phase-dependent)
power time series: amplitude time series tot de macht 2
Time-frequency plot: red is more energy, blue is less energy
- LFP is a complex and mysterious signal (lots more research to be done)
- spike sorting is also complex and mysterious
EEG data acquisition and analysis
Recording methodologies of EEG
● 10-20 system → 10 and 20% distance, odd numbers left side, even numbers right
side, A1 and A2 are reference electrodes that are placed behind ear since there
should be no electric current
, ● 15 channel SOMNO screen → home recording (for sleep eg)
● 32/64 channel Easycap/acticap + brainamp → lab recordings → there are passive and
active electrodes (active are less comfortable but better)
● 128 channel Easycap/brainamp → lab recordings
● 256 channel → too many electrodes is not always better
- caps come in all sizes (also for babies and for different head shapes)
Combinations of EEG
- EEG and fMRI → leads to a lot of artifacts and thus worse signal → measure EEG
when MRI pulse is off for better signal
- EEG and NIRS → works well in easily accessible regions from the skull
NIRS: near-infrared spectrometer → measures oxygenation of blood → better spatial
resolution than EEG
Wearable EEG caps
● lead to a lot of motion artifacts (most difficult to filter out)
- Zmax: EEG headband → limited to frontal region but you don’t have a problem with
hair
- in-ear EEG: only temporal areas, dry contact electrode (no gel needed) → dry
electrodes give worse signal than wet
Intracranial EEG and ECoG
- Depth electrodes: are the deepest and most invasive → give the best signal
- Strip electrodes
- Grid electrodes: don’t go through the skull
electrophysiology: using the electrical activity generated by the nervous system to study
neural computation, cognition, and behavior
→ cognition, emotions, etc. range from 1 to 100 Hz
How to measure neuroelectric signals
- intracellular recording: measuring from inside the neuron
- juxtacellular recording: measuring next to one specific cell in the membrane of the
brain
- extracellular recording: measuring spikes in the brain (one electrode)
- LFP/iEEG: measuring the sum of potentials in the brain (more electrodes)
- ECoG: measuring electric current on the dura (like EEG but the electrode is placed
under the skull (not in the brain))
- M/EEG: measurement outside of the head (non-invasive) → every magnetic field you
have has a perpendicular electric field
- MUA: measure collection of spikes (a little bit like LFP)
→ the further you get from the neuron, the weaker the potential that you measure with your
electrode
Advantages of electrophysiology
- cognition = fast → ephys captures this in the time-frame in which events occur (fMRI
cannot do this due to low temporal resolution) → direct measure of electrical brain
activity
- provides complex data → analyses of local regions and networks (fMRI is only one-
dimensional (up or down signal))
- link across scales/methods and species
→ species with different sizes of brains all have similar neural oscillations at almost
the same speeds (oscillations are constant across species)
Disadvantages of electrophysiology
- limited to large-scale potentials → field-cancellation might occur
- uncertainties in anatomical localization (low spatial resolution)
- data analysis is complicated, time-consuming, and annoying
- high temporal resolution → if you don’t exactly know what you are looking for you
might have a hard time finding it
How to analyze spiking data
problem: one electrode, many neurons
- use a low-pass spatial filter (LFP filter) to get local field potential
, - use spike filter → spikes contain information from many different neurons → use spike
sorting (assign each spike to a different neuron)
Traditional solution (to the aforementioned problem):
● intra/juxtacellular ephys → only measure one neuron (takes a long time to collect
data, and is ethically not doable)
Modern solution:
● spike-sorting (spike extraction → waveforms alignment → principal components
analysis → clustering (algorithm to separate spikes) → sorted waveforms)
- there are a lot of algorithms that solve this problem (hard to solve)
- spike-sorting leads to spikes going missing
, ● population → one neuron does not do exactly one thing, so we look at groups of
neurons that fire together
How to analyze LFP/MEEG data
- task-related vs. continuous
ERPs: event-related potentials → task-related analyses → take the average of all trials
- time-domain
- spectral domain
- time-frequency domain
Fourier transform: take a signal and take a sine wave as template and look at how similar
the signal is to the sine wave (phase-dependent)
power time series: amplitude time series tot de macht 2
Time-frequency plot: red is more energy, blue is less energy
- LFP is a complex and mysterious signal (lots more research to be done)
- spike sorting is also complex and mysterious
EEG data acquisition and analysis
Recording methodologies of EEG
● 10-20 system → 10 and 20% distance, odd numbers left side, even numbers right
side, A1 and A2 are reference electrodes that are placed behind ear since there
should be no electric current
, ● 15 channel SOMNO screen → home recording (for sleep eg)
● 32/64 channel Easycap/acticap + brainamp → lab recordings → there are passive and
active electrodes (active are less comfortable but better)
● 128 channel Easycap/brainamp → lab recordings
● 256 channel → too many electrodes is not always better
- caps come in all sizes (also for babies and for different head shapes)
Combinations of EEG
- EEG and fMRI → leads to a lot of artifacts and thus worse signal → measure EEG
when MRI pulse is off for better signal
- EEG and NIRS → works well in easily accessible regions from the skull
NIRS: near-infrared spectrometer → measures oxygenation of blood → better spatial
resolution than EEG
Wearable EEG caps
● lead to a lot of motion artifacts (most difficult to filter out)
- Zmax: EEG headband → limited to frontal region but you don’t have a problem with
hair
- in-ear EEG: only temporal areas, dry contact electrode (no gel needed) → dry
electrodes give worse signal than wet
Intracranial EEG and ECoG
- Depth electrodes: are the deepest and most invasive → give the best signal
- Strip electrodes
- Grid electrodes: don’t go through the skull
